blob: f78ed9d18bf0602f734682ac8f6f6c41c3c5cd26 [file] [log] [blame]
Denis Vlasenkoe4712752007-04-14 15:08:41 +00001Various bits of what is known about busybox shells, in no particular order.
2
Denis Vlasenkocccdc4e2007-11-23 21:08:38 +000032007-11-23
4hush: fixed bogus glob handling; fixed exec <"$1"; added test and echo builtins
5
Denis Vlasenkod67cef22007-06-13 06:47:47 +000062007-06-13
7hush: exec <"$1" doesn't do parameter subst
8
Denis Vlasenko163a8552007-05-24 13:52:46 +000092007-05-24
Denis Vlasenkod76c0492007-05-25 02:16:25 +000010hush: environment-related memory leak plugged, with net code size
11decrease.
12
132007-05-24
Denis Vlasenko163a8552007-05-24 13:52:46 +000014hush: '( echo ${name )' will show syntax error message, but prompt
15doesn't return (need to press <enter>). Pressing Ctrl-C, <enter>,
16'( echo ${name )' again, Ctrl-C segfaults.
17
Denis Vlasenko8805eeb2007-05-21 10:19:48 +0000182007-05-21
19hush: environment cannot be handled by libc routines as they are leaky
20(by API design and thus unfixable): hush will leak memory in this script,
21bash does not:
22pid=$$
23while true; do
24 unset t;
25 t=111111111111111111111111111111111111111111111111111111111111111111111111
26 export t
27 ps -o vsz,pid,comm | grep " $pid "
28done
29The fix is to not use setenv/putenv/unsetenv but manipulate env ourself. TODO.
30hush: meanwhile, first three command subst bugs mentioned below are fixed. :)
31
322007-05-06
Denis Vlasenko21f0d4c2007-05-06 14:15:42 +000033hush: more bugs spotted. Comparison with bash:
34bash-3.2# echo "TEST`date;echo;echo`BEST"
35TESTSun May 6 09:21:05 CEST 2007BEST [we dont strip eols]
36bash-3.2# echo "TEST`echo '$(echo ZZ)'`BEST"
37TEST$(echo ZZ)BEST [we execute inner echo]
38bash-3.2# echo "TEST`echo "'"`BEST"
39TEST'BEST [we totally mess up this one]
40bash-3.2# echo `sleep 5`
41[Ctrl-C should work, Ctrl-Z should do nothing][we totally mess up this one]
42bash-3.2# if true; then
43> [Ctrl-C]
44bash-3.2# [we re-issue "> "]
45bash-3.2# if echo `sleep 5`; then
46> true; fi [we execute sleep before "> "]
47
Denis Vlasenkoac0e5ab2007-05-04 21:37:27 +0000482007-05-04
Denis Vlasenko21f0d4c2007-05-06 14:15:42 +000049hush: made ctrl-Z/C work correctly for "while true; do true; done"
50(namely, it backgrounds/interrupts entire "while")
Denis Vlasenkoac0e5ab2007-05-04 21:37:27 +000051
Denis Vlasenkoe725bfe2007-05-03 22:45:39 +0000522007-05-03
Denis Vlasenko3349fc42007-05-04 14:54:36 +000053hush: new bug spotted: Ctrl-C on "while true; do true; done" doesn't
54work right:
55# while true; do true; done
56[1] 0 true <-- pressing Ctrl-C several times...
57[2] 0 true
58[3] 0 true
59Segmentation fault
Denis Vlasenko400c5b62007-05-04 13:07:27 +000060
612007-05-03
Denis Vlasenkoe725bfe2007-05-03 22:45:39 +000062hush: update on "sleep 1 | exit 3; echo $?" bug.
63parse_stream_outer() repeatedly calls parse_stream().
64parse_stream() is now fixed to stop on ';' in this example,
65fixing it (parse_stream_outer() will call parse_stream() 1st time,
66execute the parse tree, call parse_stream() 2nd time and execute the tree).
67But it's not the end of story.
68In more complex situations we _must_ parse way farther before executing.
69Example #2: "{ sleep 1 | exit 3; echo $?; ...few_lines... } >file".
70Because of redirection, we cannot execute 1st pipe before we parse it all.
71We probably need to learn to store $var expressions in parse tree.
72Debug printing of parse tree would be nice too.
73
Denis Vlasenko08c8c1d2007-04-28 22:39:02 +0000742007-04-28
75hush: Ctrl-C and Ctrl-Z for single NOFORK commands are working.
76Memory and other resource leaks (opendir) are not addressed
77(testcase is "rm -i" interrupted by ctrl-c).
78
Denis Vlasenko52881e92007-04-21 13:42:52 +0000792007-04-21
Denis Vlasenko1359da62007-04-21 23:27:30 +000080hush: "sleep 5 | sleep 6" + Ctrl-Z + fg seems to work.
81"rm -i" + Ctrl-C, "sleep 5" + Ctrl-Z still doesn't work
82for SH_STANDALONE case :(
83
842007-04-21
Denis Vlasenko52881e92007-04-21 13:42:52 +000085hush: fixed non-backgrounding of "sleep 1 &" and totally broken
86"sleep 1 | sleep 2 &". Noticed a bug where successive jobs
87get numbers 1,2,3 even when job #1 has exited before job# 2 is started.
88(bash reuses #1 in this case)
89
902007-04-21
Denis Vlasenko2f1bb362007-04-21 10:01:14 +000091hush: "sleep 1 | exit 3; echo $?" prints 0 because $? is substituted
92_before_ pipe gets executed!! run_list_real() already has "pipe;echo"
93parsed and handed to it for execution, so it sees "pipe"; "echo 0".
94
Denis Vlasenko54e7ffb2007-04-21 00:03:36 +0000952007-04-21
96hush: removed setsid() and made job control sort-of-sometimes-work.
97Ctrl-C in "rm -i" works now except for SH_STANDALONE case.
98"sleep 1 | exit 3" + "echo $?" works, "sleep 1 | exit 3; echo $?"
99shows exitcode 0 (should be 3). "sleep 1 | sleep 2 &" fails horribly.
100
Denis Vlasenkoe4712752007-04-14 15:08:41 +00001012007-04-14
102lash, hush: both do setsid() and as a result don't have ctty!
103Ctrl-C doesn't work for any child (try rm -i), etc...
104lash: bare ">file" doesn't create a file (hush works)